“ASR:2015-03-16”版本间的差异

来自cslt Wiki
跳转至： 导航、 搜索
（以“==Text Processing== ===LM development=== ====Domain specific LM==== * LM2.X :* train a large lm using 25w-dict.(hanzhenglong/wxx) ::* v2.0c filter the useless word....”为内容创建页面）
 
 
（某位用户的一个中间修订版本未显示）
第1行： 第1行：
 +
==Speech Processing ==
 +
=== AM development ===
 +
 +
==== Environment ====
 +
* grid-11 often shut down automatically, too slow computation speed.
 +
* GPU has being repired.--Xuewei
 +
 +
==== RNN AM====
 +
* details at http://liuc.cslt.org/pages/rnnam.html
 +
* tuning parameters on monophone NN
 +
 +
==== Mic-Array ====
 +
* reproduce environment for interspeech
 +
* investigate alpha parameter in Lasso
 +
 +
====Dropout & Maxout & rectifier ====
 +
* HOLD
 +
* Need to solve the too small learning-rate problem
 +
* 20h small scale sparse dnn with rectifier. --Mengyuan
 +
* 20h small scale sparse dnn with Maxout/rectifier based on weight-magnitude-pruning. --Mengyuan Zhao
 +
 +
====Convolutive network====
 +
* HOLD
 +
:* CNN + DNN feature fusion
 +
:* reproduce experiments -- Yiye
 +
 +
====RNN-DAE(Deep based Auto-Encode-RNN)====
 +
* HOLD -Zhiyong
 +
* http://cslt.riit.tsinghua.edu.cn/cgi-bin/cvss/cvss_request.pl?account=zhangzy&step=view_request&cvssid=261
 +
 +
====Speech rate training====
 +
:* http://cslt.riit.tsinghua.edu.cn/cgi-bin/cvss/cvss_request.pl?account=zhangzy&step=view_request&cvssid=268
 +
:* Technical report HOLD.-- Xiangyu Zeng, Shi Yin
 +
:* Paper for NCMMSC done
 +
 +
====Neural network visulization====
 +
* http://cslt.riit.tsinghua.edu.cn/cgi-bin/cvss/cvss_request.pl?account=zhangzy&step=view_request&cvssid=324
 +
* Technical report done --Mian Wang.
 +
 +
===Speaker ID=== 
 +
:* DNN-based sid --Yiye
 +
:* http://cslt.riit.tsinghua.edu.cn/cgi-bin/cvss/cvss_request.pl?account=zhangzy&step=view_request&cvssid=327
 +
 +
===Ivector based ASR===
 +
:* http://cslt.riit.tsinghua.edu.cn/cgi-bin/cvss/cvss_request.pl?step=view_request&cvssid=340
 +
:* Ivector dimention is smaller, performance is better
 +
:* Augument to hidden layer is better than input layer
 +
 +
 
==Text Processing==
 
==Text Processing==
 
===LM development===
 
===LM development===
第13行： 第62行：
 
:* error check
 
:* error check
 
* similar word extension in FST
 
* similar word extension in FST
−
:* add the experiment to tag-lm paper.
+
:* repeat the experiment using same data
  
 
====RNN LM====
 
====RNN LM====

2015年3月18日 (三) 07:27的最后版本

Speech Processing

AM development

Environment

  • grid-11 often shut down automatically, too slow computation speed.
  • GPU has being repired.--Xuewei

RNN AM

Mic-Array

  • reproduce environment for interspeech
  • investigate alpha parameter in Lasso

Dropout & Maxout & rectifier

  • HOLD
  • Need to solve the too small learning-rate problem
  • 20h small scale sparse dnn with rectifier. --Mengyuan
  • 20h small scale sparse dnn with Maxout/rectifier based on weight-magnitude-pruning. --Mengyuan Zhao

Convolutive network

  • HOLD
  • CNN + DNN feature fusion
  • reproduce experiments -- Yiye

RNN-DAE(Deep based Auto-Encode-RNN)

Speech rate training

Neural network visulization

Speaker ID

Ivector based ASR


Text Processing

LM development

Domain specific LM

  • LM2.X
  • train a large lm using 25w-dict.(hanzhenglong/wxx)
  • v2.0c filter the useless word.(next week)
  • set the test set for new word (hold)
  • prepare the wiki data: entity list.

tag LM

  • Tag Lm(JT)
  • error check
  • similar word extension in FST
  • repeat the experiment using same data

RNN LM

  • rnn
  • the input and output is word embedding and add some token information like NER..
  • map the word to character and train the lm.
  • lstm+rnn
  • check the lstm-rnnlm code about how to Initialize and update learning rate.(hold)

Word2Vector

W2V based doc classification

  • data prepare.(hold)

Knowledge vector

  • make a report on Monday

Translation

  • v5.0 demo released
  • cut the dict and use new segment-tool

Sparse NN in NLP

  • prepare the ACL
  • check the code to find the problem .
  • increase the dimension
  • use different test set.

QA

improve fuzzy match

  • add Synonyms similarity using MERT-4 method(hold)

online learning

  • data is ready.prepare the ACL paper
  • prepare sougouQ data and test it using current online learning method

framework

  • extract the module
  • extract the context module ,search module,entity recognize module and common module.
  • define the inference in different modules
  • composite module

leftover problem

  • new inter will install SEMPRE
取自“http://cslt.org/mediawiki/index.php?title=ASR:2015-03-16&oldid=14286”